Qwen2.5-3B + GRPO
| Method | Links | |||||
|---|---|---|---|---|---|---|
2026.03 | 60.7 | 63.6 | 79 | 33.6 | 85.1 | |
2026.03 | 58.5 | 61.2 | 75.6 | 33.3 | 78.3 | |
2026.03 | 57.2 | 61.6 | 76.9 | 30.7 | 78 | |
2026.03 | 41.5 | 47.9 | 70.6 | 18.2 | 57.8 | |
2026.03 | 38.6 | 45.8 | 69.7 | 17 | 54.1 | |
2026.03 | 38.4 | 45.7 | 69.7 | 17 | 54.1 | |
2026.03 | 38.1 | 45.6 | 69.6 | 16.8 | 51.7 | |
2026.03 | 37.6 | 45.1 | 69.5 | 15.4 | 50.5 | |
2026.03 | 37.5 | 45.1 | 69.3 | 16 | 51.5 |